How Copilot/Vibe Coding/Agentic A.I are slowly taking away developer jobs (Telugu)

Brief Summary

This video discusses the potential drawbacks of over-reliance on AI coding assistants like GitHub Copilot, particularly the "agent mode," which automates code changes. It highlights the risk of developers losing their understanding of the codebase and the reasoning behind changes, potentially leading to negative consequences such as job loss and difficulty in technical discussions. The speaker emphasizes the importance of reviewing and understanding AI-generated code changes to maintain technical proficiency and avoid introducing errors.

  • Over-reliance on AI coding assistants can hinder developers' understanding of the codebase.
  • The "agent mode" in tools like Copilot automates code changes, potentially leading to a lack of review and understanding.
  • Failure to understand and verify AI-generated code can result in errors, job loss, and difficulty in technical discussions.
  • It's crucial to review and understand AI-generated code changes to maintain technical proficiency.

Introduction: AI Coding Assistants and Their Impact

The speaker addresses the increasing use of AI coding assistants like GitHub Copilot in software development. While these tools can boost productivity by quickly providing solutions to coding problems, there's a potential downside. The speaker warns against blindly trusting AI-generated code, especially with the advent of "agent mode," which automatically implements changes in the codebase.

The Problem with "Agent Mode": Loss of Understanding

The speaker contrasts the traditional "ask and edit" mode with the newer "agent mode." In the former, developers would research solutions, understand the reasoning behind changes, and manually implement them. With "agent mode," the AI analyzes the code base, creates files, and makes changes automatically, requiring only approval from the developer. This automation can lead to developers not fully understanding the changes being made, which can be detrimental in the long run.

Consequences of Over-Reliance: Technical Skills and Job Security

The speaker emphasizes that a strong technical foundation is crucial for career advancement and job security. Relying solely on AI to generate code can hinder the development of these skills. The speaker shares an anecdote about a development team where five members were replaced after it was discovered they were blindly accepting AI-generated code changes without understanding their impact. This resulted in errors and broken flows in the application, ultimately leading to their dismissal.

The Importance of Reviewing AI-Generated Code

The speaker concludes by stressing that using AI coding assistants is not inherently wrong, but it's essential to use them as supportive tools rather than replacements for critical thinking. Developers should always review the changes made by AI, understand the reasoning behind them, and ensure they don't introduce errors or break existing functionality. Being able to explain the code and the changes made is crucial for maintaining credibility and job security.

Summarize any YouTube video in seconds

Get summaries like this one for any video, in your language, with the free Summ app.

Download on the Apple Store
Get it on Google Play

Transcript

Full transcript of “How Copilot/Vibe Coding/Agentic A.I are slowly taking away developer jobs (Telugu)”. Click a timestamp to open the video at that moment. Need another video? Use the YouTube transcript generator.

0:00 Friends, don't use AI, don't put your code in chat, because it will go into the open source public. The companies that said that our company is specific proprietary code are now hosting things like CoPilot by CoPilot in the company. And developers are truly encouraged to use CoPilot for their work. In fact, using CoPilot increases productivity and it is true that once we had an issue like Stack Overflow, Google, etc., it would take around one or two hours to finally understand the issue. After CoPilot came, it would hardly take you five minutes to find the solution. But here is where the problem is. Friends, once the only ask and edit mode in EAI It used to be that if we

0:46 asked a question, it would provide related information documents and a solution to that question. So we will try that solution. Before implementing that solution, we will understand why we need to change it. We used to ask at least why we are going to go and understand this solution. And also, that solution explains what needs to be changed. We used to say how this particular change will impact us and where we need to make changes. But recently, a mode called Agent has come up. Friends, what is this mode? It means that we don't need to make any changes. If we just give our question and say, "You make all these changes in the code base," it automatically analyzes the entire workspace, the entire code base, whatever we give it as input, and it also

1:33 goes and creates the files that need to be created. The only thing we have to do is see the changes made and press OK, Discard OK, that's it. All changes will be there. Here is where the problem is. Even if we want to change jobs in the future or go to future technical discussions, we have to understand like we have to be technically strong. Even if we take our time on Stack Overflow and Google, we know multiple solutions by reading multiple resources, we understand multiple concepts, but when you go to the agent mode, we all know that we don't even need to see what is actually happening there. The deadlines in our companies are very tight, and we have to complete the work fast . And with copilot provided to the developers, the deadlines are getting even

2:19 stricter. I also saw in my surroundings that this copilot agent didn't even look at what we said since he arrived . We are putting it in agent mode, asking a question, it creates a solution, we are pressing accept. In the short term, all the timelines are met, all the work is completed, but in the long term, what do we do with those concepts? After going into the interview, if we are asked in-depth questions and do not know how to make a specific change, then it is going to be a problem in our interviews and in our future. Also, recently, in my circle, I saw that five people were removed from a 25-member development team in the last two months. Was it because of lack of budget or was it because of reducing resources? I mean, what is it? They removed those five people and hired another five within a week .

3:05 If we do a little analysis of what actually happened, when we spoke to those five people, they used GitHub Copilot extensively. The company has provided GitHub CoPilot. CoPilot has an ask mode, an edit mode, and an agent mode. It's a good idea to use what the company provided . Otherwise, there used to be only one Ask mode. In Ask, we ask a question, it gives us an answer, we see why, we make changes. In this agent mode, we ask questions. That doesn't answer. It implements the solution as well as the answer . It goes into our workspace and makes all the code changes . It's fine as long as it works . But in the last one or two months, they've been saying, "Hey, they're making the same changes, we just have to click accept." We're looking at the above and wondering why we're even bothering, but when we make a particular change in the code base of such a large company, we have to see if it breaks any of the adjacent flows, not just the flow we're testing. They saw nothing and other

3:52 threads started breaking. When the client analyzed when other threads started breaking, they looked at where these issues were coming from, and this particular code came up. If you look at who are developers, there are these five . These five people came up to me and didn't even ask me, as a normal client, why we implemented these particular changes like this or why the side threads broke. When asked why we implemented these particular changes this way, these developers said, "We don't know. We prompted the angry pilot to come up with a solution that we wanted, and the angry pilot made those changes." Sometimes, however, they don't even know the file is there . When they talked to their client about seeing that file for the first time, they asked why they created this file, saying that it was n't necessary, that the co- pilot created it. Every time they did this, it became clear that they had removed these five people from the project and

4:37 replaced them with five others. So friends, the only thing we need to know from this is that using Copilot is not wrong. The company gives us it. It should be a supportive agent. If we go into agent mode and do not review the changes that are happening there completely, in the future, not only with the company, when we want to change jobs, we will end up in a situation where we can't explain it. So if you are using Copilot, in fact, I am also using Young Copilot, but in that agent mode, at least in every file, what changes have been made? Is that change really necessary or not? It is extremely important if you are using any Copilot agent, please review the changes and understand why that change was actually

5:25 made. It is important because when the client asks Either when our manager and lead asked, it is truly not acceptable to say that the pilot did this, it is like this, it is not acceptable to say that we verify it and then we have to put

Share

More video summaries